BY THE END OF 2006, 40 million people worldwide were living with human immunodeficiency virus (HIV), and the majority of these, 25 million, reside in sub-Saharan Africa.1 Without effective measures to treat the disease and address its spread, countries in the region face a demographic and economic catastrophe of staggering dimensions as evidenced by increasing acquired immunodeficiency virus (AIDS)-related morbidity and mortality and decreasing life expectancy at birth. The World Health Organization estimates that the current life expectancy in the sub-Saharan region of 43 years at birth would be 56 years in the absence of the epidemic, and that 1 in 5 deaths annually in Africa is attributable to acquired immunodeficiency virus.2,3
Moreover, despite stabilizing and declining trends in some countries, notably Uganda, the HIV epidemic is expanding rapidly in most parts of the region. Southern Africa is the worst affected area, with prevalence of HIV among sexually active adults ranging between 20% and 40% in most countries where the epidemic is generalized.1 The prevalence of infection among adults is typically around 10% in countries of Central African region, and 5% or higher in 3 West African countries of Bukina Faso, Cote d’Ivore, and Nigeria.
The search for effective interventions to reduce further expansion of the epidemic and mitigate its social and economic impact continues to be therefore the most important public health priority in sub-Saharan Africa. To this end, improved understanding of the behavioral risk factors that amplify the force of the infection is critical in the design of effective interventions and in targeting interventions to people at highest risk of infection.
One behavior that may be implicated in the transmission of HIV and other sexually transmitted diseases (STDs) is alcohol consumption. A recent systematic review of published literature found problematic alcohol consumption was associated with increased risk of STDs.4 The contribution of alcohol drinking to HIV transmission has not been systematically examined, and most intervention programs neglect the potential role of alcohol consumption as an intermediate variable, which may influence behaviors related to HIV. To address these omissions and to quantify the association between alcohol drinking behavior and HIV infection, we conducted a systematic review of the scientific literature and a meta-analysis of the reported risks associated with alcohol use and HIV based on studies conducted in Africa.
Figure 1 summarizes the process for selecting articles included in the meta-analysis. Initially, EMBASE and PubMed databases were searched to identify studies conducted in Africa that related alcohol use to HIV infection. To expand the scope of the enumeration the search was repeated for alcohol use and sexually transmitted infections (STIs) and again for alcohol and STDs. In all the 5 separate searches identified 250 unduplicated references, 96 from PubMed and 66 from EMBASE uniquely, and 88 that appeared in both databases.
Titles and abstracts for the unduplicated references were reviewed and 77 of 250 articles were retained after meeting the following criteria: 1) reported original, empirical research published in a peer review journal, and 2) considered alcohol use or drinking behavior as a potential risk factor for HIV infection. Studies of HIV risk factors such as presence of other STIs, high-risk sexual behavior, and sexual violence were included in the review in case any reported rates of HIV infection among drinkers and nondrinkers.
Out of 77 studies reviewed, 20 provided data that could be used in the meta-analysis. These studies either reported odds ratios (ORs) comparing drinkers and nondrinkers about HIV prevalence and/or seroconversion, or a contingency table from which the ORs could be calculated. A list of the studies used in the analysis and a digest of the information extracted from them is given in Table 1.
Eighteen of the studies are cross-sectional and hence measured the relationship between alcohol use and HIV prevalence in the sample, whereas 25,6 are prospective and examined HIV transmission among drinkers and nondrinkers. Additionally, 2 types of samples were encountered: representative samples from community-based surveys5,7–14 and samples of high-risk groups such as bar and hotel workers,15,16 sailors,17 miners,14 beer hall patrons,18 pregnant women and mothers,6,19,20 and people seeking treatment in health facilities.21–24 Perhaps most significantly, all the studies represented were conducted in East Africa and Southern Africa.
Three studies in addition to those listed in Table 1, although they did not report data that could be used in the meta-analysis, provided evidence of a positive relationship between alcohol use and HIV status, risk, or transmission. These include an investigation of HIV prevalence and high alcohol consumption levels among female recreation workers in Tanzania,25 an ecological analysis in Tanzania relating local alcohol use with HIV prevalence rates,26 and a Rwandan study that associated HIV status of a woman with alcohol use of her partner.27
Measures of alcohol use were not consistent across studies. Drinking was indicated by quantity and frequency consumed, standardized diagnostic instrument of problem drinking (e.g., CAGE, AUDIT), or by attending bars or selling alcohol at home. The time period measured also was not uniform, some studies asked about recent drinking, past week as an example, whereas others used an ever use/never use dichotomy. For purposes of the analysis then the criterion was alcohol use, yes or no, regardless of the time period covered and the outcome was the binary variable HIV status, positive or negative.
A second reader independently verified the extracted data. The data obtained were quite simple and no standardized coding form was needed or used. The coders were not masked to the journals or authors of the studies reviewed. There were no conflicts between the coders in the classification of the extracted data.
The meta-analysis was guided by the checklist proposed by the Meta-analysis of Observational Studies in Epidemiology (MOOSE) Group.28 Weighted summary measures were computed as ORs with 95% confidence intervals (CIs), calculated by the DerSimonian and Laird’s random-effects model,29 where OR was chosen to make the statistical measure for each study consistent. Evidence of publication bias was investigated by means of funnel plot and 2 different tests proposed by Begg and Mazumdar30 based on rank correlation (after continuity correction) and by Egger et al.31 Heterogeneity was quantitatively assessed using the Cochran’s Q statistics and also visually examined through the L’Abbe plot32 (not shown).
Meta-analyses of univariate ORs and multivariate adjusted odds ratios (AORs) were performed and forest plots were used to visualize individual as well as the pooled effects. As a sensitivity analysis, we implemented the trim and fill method33 to account for publication bias that was hidden or undetected by standard statistical test.
To address the effects of important study characteristics, we additionally conducted several subgroup analyses, by: sex of subjects in the sample (male vs. females) and type of sample (population based vs. high risk).
The primary analysis combined studies having disparate definitions of alcohol consumption. To further specify the relationship of interest the analysis was repeated with the subset of studies (7) that rigorously defined alcohol use and differentiated between problem and nonproblem drinkers. Standard definitions of problem drinking used to categorize drinkers were: quantity, ≥3 drinks per occasion;23 frequency, daily consumption;9,11,14,16 and diagnostic instrument scores (CAGE, AUDIT).15,22
A 2-sided hypothesis was employed in all statistical testing and CI construction. Statistical and graphical analyses were carried out by STATA version 8.2.34
The results of all the analyses are summarized in Table 2, while Figure 2 displays the primary analyses using crude ORs. Univariate analysis of the association between alcohol consumption and HIV infection shows that alcohol users have significantly higher odds of being HIV-positive (pooled OR = 1.70; 95% CI = 1.45–1.99; P value <0.001). The test of heterogeneity was highly significant with P value <0.001, whereas publication bias was not detected (P value = 0.67 and 0.12 by Begg and Mazumdar’s and Egger et al.’s tests, respectively).30,31
Next, as a confirmatory measure, we repeated the analyses implementing a trim and fill methodology that accounts for potential publication bias that might not have been evident in our original investigation but still might exist. This method rank-orders the ORs and determines from the gaps in the distribution the number of studies that may potentially have been excluded because of publication bias. Hypothetical results are created to correct for these possible omissions and the pooled OR is computed on the larger sample of studies. Applying this method to the univariate ORs, 5 negative studies were estimated to be missing and results for them were created artificially. Based on analysis of this expanded study sample, alcohol drinkers remained at relatively increased risk when compared with nondrinkers (OR = 1.48; 95% CI = 1.26–1.73; P value <0.001).
The same analysis was repeated among 11 studies that reported multivariable AORs for the association between alcohol use and HIV infection; the results are summarized in Table 2 and Figure 3. In this analysis the association between HIV infection and alcohol use was attenuated to OR = 1.57 (95% CI = 1.42–1.72; P value <0.001). Heterogeneity among studies was no longer significant (P = 0.60) and again publication bias was not significant (P = 0.34–0.43). After applying the trim and fill methodology and filling 3 studies that could have been omitted, the results were largely unchanged (OR = 1.52; 95% CI = 1.39–1.66; P value < 0.001).
Funnel plots that aid assessment of publication bias are presented in Figure 4.
To examine what effect study characteristics might have on the results, we analyzed subgroups of studies defined by sample characteristics, by gender and if the sample was high risk or population based (Table 2). About the former when we first compared all studies with homogenous samples of either males (10 studies) or females (16 studies) the magnitudes of association were nearly identical, OR = 1.91 (95% CI = 1.57–2.33) and OR = 1.90 (95% CI = 1.68–2.19), respectively.
We repeated this analysis restricting the sample to studies that reported results for both males and females. The same study restriction tends to eliminate potential sources of bias such as geography, time, and alcohol measurement. Analysis of the 8 studies that fit this criterion produced pooled ORs that were more discrepant but still not meaningfully different: males OR = 1.98 (95% CI = 1.76–2.22) and females OR = 1.78 (95% CI = 1.53–2.08).
Alcohol use was associated with significantly higher odds of being HIV-positive for both types of samples, population based (OR = 1.77; 95% CI = 1.62–1.93) and high-risk groups (OR = 2.01; 95% CI = 1.56–2.58), where the latter group showed a stronger association as anticipated (z = 1.91, P <0.06).
Problem drinkers were at greater risk to be HIV+ (Table 2) than were drinkers who did not use alcohol symptomatically (z = 2.08, P <0.04). When compared with nondrinkers, the pooled estimates of HIV risk was 1.57 (95% CI = 1.33–1.86) for nonproblem alcohol drinkers and 2.04 (95% CI = 1.61–2.58) for problem drinkers.
We identified 20 studies that examined the association between alcohol consumption and HIV infection in Africa. Based on a systematic review and meta-analysis, we observed evidence of a strong relationship between alcohol use and HIV infection. Drinkers have a 70% greater chance of being HIV positive when compared with nondrinkers in the bivariate case, and a 57% increased risk of HIV infection when potential confounders were controlled in multivariate analysis. The association remained strong and consistent after adjustment for possible publication bias with elevated risks of alcohol use ranging from 48% to 52% for the univariate and multivariate pooled risk estimates respectively.
This conclusion is further supported by 3 additional findings of the analysis. First, as shown in the plots of the effects by study, there is strong consistency and agreement in the estimated effect measures across studies, particularly for studies reporting a multivariate AOR. Second, risk estimates are comparable across different types of samples, for males and females and for high-risk and representative populations. Third, alcohol use demonstrates a crude dose–response relationship with HIV infection such that the heaviest and symptomatic drinkers are at greater risk to be HIV+ than are more moderate drinkers and those who do not experience problems as a result of drinking. Taken together, these results provide strong evidence supporting the role of alcohol drinking as a risk factor for HIV infection.
These findings and conclusions should be interpreted in light of a number of limitations of our analysis. With exception of 2 reports all studies we examined were cross-sectional and it is not possible to determine if a causal relationship between alcohol use and HIV infection exists. An equally plausible explanation for some or all of the observed association is that people living with HIV drink more to cope with their circumstance. Results from the 2 prospective studies conducted in Uganda5 and Tanzania6 did show, however, that drinking was strongly related to increased risk of HIV acquisition (incidence rate ratio = 1.89; 95% CI = 1.59–2.26 and incidence rate ratio = 2.61; 95% CI = 1.63–4.16, respectively). In both studies the relationship remained after adjustment for other factors in multivariate regression analysis.
Every meta-analysis is subject to publication bias because nonsignificant findings are less apt to make their way into print.35 This problem is exacerbated when statistical models are employed because customarily only significant terms are reported. In our case nonsignificant results from at least 3 studies7,9,22 could not be included in the multivariate meta-analysis for this reason. To partially remedy the problem, we adopted a trim and fill methodology to simulate the effect of missing studies on the pooled estimates. The association between alcohol use and HIV status may be overstated nonetheless.
It is important to note that significant heterogeneity exists in the unadjusted analysis. Given this situation combining the effects may not be justified. However, we also note that the adjusted analysis, subgroup and problem drinker analyses did not exhibit the heterogeneity observed among the univariate ORs. Results of these analyses gave some insight into the source of the heterogeneity among the univariate ORs, therefore and suggest that our pooled estimates based on adjusted and subgroup analyses are not subject to the same caveat.
We have only examined the potential direct effect of alcohol use on HIV status. As discussed further in this report, alcohol use could be associated with HIV indirectly through other risk factors such as high-risk sexual behavior, gender violence, and presence of other STDs. Detailed analysis to investigate the effects of alcohol use on these potential intermediate variables could shed more light on this complex relationship between alcohol use and HIV infection.
Finally, the analysis suffers from the vagaries in the way alcohol use was measured across the studies. This had the immediate effect of forcing aggregation of alcohol use into a very crude yes/no dichotomy. In the long run, such categorization works against an in-depth understanding of the relationship between alcohol use and HIV infection. The knowledge base will be advanced significantly if standardized measures of alcohol use and abuse are adopted in future studies.
There are several potential mechanisms that may explain the observed association between alcohol consumption and increased risk of HIV. In most African countries, alcohol is consumed in small bars and other informal alcohol-serving establishments patronized by high-risk men seeking new sex partners.18,36 As a central nervous depressant, alcohol increases sexual risk taking by reducing inhibitions and diminishing perception of exposure risks.11 Thus, the social context prevailing in these places combined with alcohol use may contribute in increased transmission of HIV by facilitating high-risk encounters.
HIV risk factors and high-risk sexual behaviors have been shown to be related to drinking as well. Alcohol use has as an example been associated with higher rates of unprotected intercourse,21,37 failure to use condoms appropriately,21 and increased frequency of sexual activity and/or number of sexual partners.21,38,39 Other demonstrated associations include gender violence, sexual assault, and rape11,21,40,41 factors that are known to increase women’s vulnerability to HIV, and susceptibility to other STDs4 which is known to be associated with increased risk of HIV.42 Furthermore, there is a growing body of evidence that implicates chronic alcohol ingestion with compromised immune response43,44 and increased susceptibility to HIV infection.45–47
Overall this review of the published literature confirms that alcohol is an important risk factor for HIV infection in Africa. The accumulated evidence suggests additionally that prevention efforts should focus on groups most likely to experience alcohol problems, such as frequent, heavy, and symptomatic drinkers, and needs to take into account the social context in places where alcohol is consumed. In this regard, modification and implementation of extant simple and efficacious methods that are graduated to increasing involvement with alcohol, e.g., Brief Interventions developed by the WHO,48 could be viable alternatives.
Levels of alcohol consumption in many African countries are increasing, mainly because of the changing drinking patterns from low alcohol content traditional home-brews to recreational use of high alcohol content commercial brews49–51 and increased availability and accessibility of alcoholic beverages.51 This trend and the strength of the association between drinking and HIV status observed in this study suggest that alcohol use may amplify the current and future course of the epidemic. Thus, effective interventions directed toward this modifiable behavior may have substantial impact in reducing the devastating impact of the epidemic in this region of the world.
1. UNAIDS AIDS epidemic update, December 2006. Available at http://www.unaids.org/en/HIV_data/epi2006/default.asp
. Accessed March 2, 2007.
2. World Health Organization. The World Health Report 2004: Changing History. Geneva, Switzerland: World Health Organization, 2005.
3. World Health Organization. The World Health Report 2003, Annex Table 2. Geneva, Switzerland: World Health Organization, 2003:154.
4. Cook RL, Clark DB. Is there an association between alcohol consumption and sexually transmitted diseases? A systematic review. Sex Transm Dis 2005; 32:156–164.
5. Zablotska IB, Gray RH, Sewadda D, et al. Alcohol use before sex and HIV acquisition: A longitudinal study in Rakai, Uganda. AIDS 2006; 20:1191–1196.
6. Kapiga SH, Lyamuya EF, Lwihula GK, et al. The incidence of HIV infection among women using family planning methods in Dar es Salaam, Tanzania. AIDS 1998; 12:75–84.
7. Kapiga SH, Sam NE, Mlay J, et al. The epidemiology of HIV-1 infection in northern Tanzania: Results from a community-based study. AIDS Care 2006; 18: 379–387.
8. Lewis JJC, Garnett GP, Mhlanga S, et al. Beerhalls as a focus for HIV prevention: Activities in rural Zimbabwe. Sex Transm Dis 2005; 32:364–369.
9. Zuma K, Gouws E, Williams B, et al. Risk factors for HIV infection among women in Carletonville, South Africa: Migration, demography and sexually transmitted diseases. Int J STD AIDS 2003; 14:814–817.
10. Hargreaves JR, Morison LA, Chege J, et al. Socioeconomic status and risk factors of HIV infection in an urban population in Kenya. Trop Med Int Health 2002; 7:793–802.
11. Mbulaiteye SM, Ruberantwari A, Nakiyingi JS, et al. Alcohol and HIV: A study among sexually active adults in rural southwest Uganda. Int J Epidemiol 2000; 29:911–915.
12. Tengia-Kessy A, Msamanga GI, Moshiro CS. Assessment of behavioral risk factors associated with HIV infection among youth in Moshi rural district, Tanzania. East Afr Med J 1998; 75:528–532.
13. Mnyika KS, Klepp KI, Kvale G, et al. Risk factors for HIV-1 infection among women in the Arusha region of Tanzania. J Acquir Immune Defic Syndr Hum Retrovirol 1996; 11:484–491.
14. Gilgen D, Williams BG, Campbell CM. HIV/AIDS in Carletonville: The 1998 Survey. Johannesburg: CSIR, 2000.
15. Ao TH, Sam NE, Masenga EJ, et al. HIV-1 among bar and hotel workers in Tanzania: The interplay between alcohol, HSV-2 and HIV-1 infection. Sex Transm Dis 2006; 33:163–169.
16. Kapiga SH, Sam NE, Shao JF, et al. HIV-1 epidemic among female bar and hotel workers in northern Tanzania: Risk factors and opportunities for prevention. J Acquir Immune Defic Syndr 2002; 29:409–417.
17. Demissie K, Amre D, Tsega E. HIV-1 infection in relation to educational status, use of hypodermic injections and other risk behaviors in Ethiopian sailors. East Afr Med J 1996; 73:819–822.
18. Fritz KE, Woelk GB, Bassett MT, et al. The association between alcohol use, sexual risk behavior, and HIV infection among men attending beerhalls in Harare, Zimbabwe. AIDS Behav 2002; 6:221–228.
19. Ayisi G., van Eijk AM, ter Kuile FO, et al. Risk factors for HIV infection among asymptomatic pregnant women attending an antenatal clinic in western Kenya. Int J STD AIDS 2000; 11:393–401.
20. Braddick MR, Kreiss JK, Embree JB, et al. Impact of maternal HIV infection on obstetrical and neonatal outcome. AIDS 1999; 4:1001–1005.
21. Simbayi LC, Kalichman SC, Jooste S, et al. Alcohol use and sexual risks for HIV infection among men and women receiving sexually transmitted infection clinic services in Cape Town, South Africa. J Stud Alcohol 2004; 65:434–442.
22. Sebit MB, Tombe M, Siziya S, et al. Prevalence of HIV/AIDS and psychiatric disorders and their related risk factors among adults in Epworth, Zimbabwe. East Afr Med J 2003; 80:503–512.
23. Talbot EA, Kenyon TA, Moeti TL, et al. HIV risk factors among patients with tuberculosis-Botswana 1999. Int J STD AIDS 2002; 13:311–317.
24. Ndinya-Achola JO, Ghee AE, Kihara AN, et al. High HIV prevalence, low condom use and gender differences in sexual behavior among patients with STD-related complaints at a Nairobi primary health care clinic. Int J STD AIDS 1997; 8:506–515.
25. Clift S, Anemona A, Watson-Jones D, et al. Variations of HIV and STI prevalences within communities neighboring new goldmines in Tanzania: Importance for intervention design. Sex Transm Infect 2003; 79:307–312.
26. Boerma JT, Urassa M, Senkoro K, et al. Spread of HIV infection in a rural area of Tanzania. AIDS 1999; 13:1233–1240.
27. Allen S, Lindan C, Serufilira A, et al. Human immunodeficiency virus infection in rural Rwanda. JAMA 1991; 266:1657–1663.
28. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology. A proposal for reporting. JAMA 2000; 283:2008–2012.
29. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177–188.
30. Begg CB, Mazumda M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994; 50:1088–1101.
31. Egger M, Smith GD, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315:629–634.
32. L’Abbe KA, Detsky AS, O’Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987; 107:224–233.
33. Duval SJ, Tweedie RL. Trim and fill: A simple funnel plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2000; 56:276–284.
34. Stata [computer program] Version 8.2 Intercooled. College Station, TX: Stata Corporation, 2005.
35. Young SS, Bang H. The file-drawer problem, revisited [Letters to the Editor]. Science 2004; 306:1133–1134.
36. Weir SS, Pailman C, Mahlalela X, et al. From people to places: Focusing AIDS prevention efforts where it matters most. AIDS 2003; 17:895–903.
37. Seeley JA, Malamba SS, Nunn AJ, et al. Socioeconomic status, gender and risk of HIV-1 infection in a rural-community in south-west Uganda. Med Anthropol Q 1994; 8:78–79.
38. Leigh BC, Temple MT, Trocki FK. The relationship of alcohol use to sexual activity in a US national sample. Soc Sci Med 1994; 39:1527–1535.
39. Somose P, Chapko MK, Hawkins RV. Multiple sexual partners: Results of national HIV/AIDS survey in the Central African Republic. AIDS 1993; 7:579–583.
40. Kalichman SC, Simbayi LC. Sexual assault history and risks for sexually transmitted infections among women in Cape Town, South Africa. AIDS Care 2004; 16:681–689.
41. Koenig MA, Lutalo T, Zhao F, et al. Coercive sex in rural Uganda: Prevalence and associated risk factors. Soc Sci Med 2004; 58:787–798.
42. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: The contribution of other sexually transmitted diseases to sexual transmission of HIV-1 infection. Sex Transm Infect 1999; 75:3–17.
43. Kovacs EJ, Messingham KA. Influence of alcohol and gender on immune response. Alcohol Res Health 2002; 26:257–263.
44. Szabo G. Alcohol’s contribution to compromised immunity. Alcohol Health Res World 1997; 21:30–41.
45. Bagasra O, Kajdacsy-Balla A, Lischner HW. Effects of alcohol ingestion on in vitro susceptibility of peripheral blood mononuclear cells to infection with HIV and of selected T-cell functions. Alcohol Clin Exp Res 1989; 13:636–643.
46. Bagasra O, Kajdacsy-Balla A, Lischner HW, et al. Alcohol intake increases human immunodeficiency virus type 1 replication in human peripheral blood mononuclear cells. J Infect Dis 1993; 167:789–797.
47. Bagasra O, Whittle P, Kajdacsy-Balla A, et al. Effects of alcohol ingestion on in vitro susceptibility of peripheral blood mononuclear cells to infection with HIV-1 and on CD4 and CD8 lymphocytes. Prog Clin Biol Res 1990; 325:351–358.
48. Barbor TF, Higgins-Biddle JC. Brief intervention for hazardous and harmful drinking: A manual for use in primary care. Geneva, Switzerland: World Health Organization, 2001. Available at http://whqlibdoc.who.int/hq/2001/WHO_MSD_MSB_01.6b.pdf
. Accessed March 15, 2007.
49. Abiodun OA. Alcohol-related problems in primary care patients in Nigeria. Acta Psychiatr Scand 1996; 93:235–239.
50. Kebede D, Alem A. The epidemiology of alcohol dependence and problem drinking in Addis Ababa, Ethiopia. Acta Psychiatr Scand 1999; 397(suppl):30–34.
51. Parry CDH, Pluddemann A, Steyn K, et al. Alcohol use in South Africa: Findings from the demographic and health survey (1998). J Stud Alcohol 2005; 66:91–97.